Speak for the Trees

Extracts decision rules from tree- and rule-based models fitted in 'R'. Rules are expressed as logical predicates that identify paths to terminal nodes, making model behavior more transparent and interpretable. Provides conversion methods to 'partykit' party objects for a wide range of model types. The 'partykit' infrastructure is described in Hothorn and Zeileis (2015) < https://jmlr.org/papers/v16/hothorn15a.html>.

lorax

The goal of lorax is to help look at different aspects of tree- and rule-based models.

lorax supports a few APIs:

• as.party() converts trees to the format used by partykit::ctree(), mostly because it has an amazing plot() method.
• extract_rules() helps write out the logical paths to the terminal nodes.
• active_predictors() enumerates which predictors were actually used in a split.
• var_imp() is a wrapper for any importance method contained in the package. This is a wrapper that enables a common interface to the scores.

Here is a list of which classes have which methods:

Note that as.party.rpart() is in the partykit package and that cforest is made out of party objects.

Installation

You can install the released version of lorax from CRAN:

install.packages("lorax")

Or install the development version from GitHub:

pak::pak("tidymodels/lorax")

Example

set.seed(822)
rngr_fit <- ranger(species ~ ., data = penguins, max.depth = 3, num.trees = 10)

rngr_party <- as.party(rngr_fit, tree = 1, data = penguins)
rngr_party
#> 
#> Model formula:
#> ~island + bill_length_mm + bill_depth_mm + flipper_length_mm + 
#>     body_mass_g + sex + year
#> 
#> Fitted party:
#> [1] root
#> |   [2] flipper_length_mm <= 206.5
#> |   |   [3] body_mass_g <= 2800: Adelie (n = 3, err = 66.7%)
#> |   |   [4] body_mass_g > 2800
#> |   |   |   [5] bill_length_mm <= 44.15: Adelie (n = 152, err = 3.9%)
#> |   |   |   [6] bill_length_mm > 44.15: Chinstrap (n = 60, err = 6.7%)
#> |   [7] flipper_length_mm > 206.5
#> |   |   [8] bill_depth_mm <= 18.15: Gentoo (n = 123, err = 0.8%)
#> |   |   [9] bill_depth_mm > 18.15: Chinstrap (n = 6, err = 16.7%)
#> 
#> Number of inner nodes:    4
#> Number of terminal nodes: 5

plot(rngr_party)

all_rules <- extract_rules(rngr_party, trees = 10)

# An expression
all_rules$rules[[1]]
#> flipper_length_mm <= 206.5 & body_mass_g <= 2800

# Text
all_rules$rules[[1]] |> rule_text()
#> [1] "flipper_length_mm <= 206.5 & body_mass_g <= 2800"

# Substitutions
new_names <- 
 tribble(
  ~ original, ~ label,
  "flipper_length_mm", "Flipper Length",
  "body_mass_g", "Body Mass"
 )
 
all_rules$rules[[1]] |> rule_text(key = new_names)
#> [1] "Flipper Length <= 206.5 & Body Mass <= 2800"

# Bullets:
all_rules$rules[[1]] |> 
 rule_text(key = new_names, bullets = TRUE) |> 
 cat()
#> * Flipper Length <= 206.5
#> * Body Mass <= 2800

Code of Conduct

Please note that the lorax project is released with a Contributor Code of Conduct. By contributing to this project, you agree to abide by its terms.